Radiology, AI, and the Fear of Obsolescence
Over the past few weeks, I’ve received two emails that prompted deep reflection. The first came from a father concerned that his medical student daughter is considering radiology. He worries that AI will eventually replace radiologists, making it a poor career choice. The second was from a radiology resident who described constant discouragement—from peers in other specialties who tell him he’ll soon need another career plan, and from parents actively urging him to switch fields in medicine. Radiology, he wrote, is his true passion. He doesn’t want to leave it; he wants to excel in it—but he wants to know what I think.
Both messages center on the same question: should young physicians—or those soon to enter medicine—pursue radiology? And should current radiologists and trainees heed the well-intentioned fears of those around them, who don’t want to see years of effort end in obsolescence?
After reflection, and conversations with colleagues in my own practice, I’ve come to several conclusions.
First, we must all accept an uncomfortable truth: none of us can predict the future—though many confidently act as if they can. Just a few weeks ago, while riding a stationary bike and watching YouTube talks about the societal impacts of AI (not radiology specifically), one well-known international author casually dropped this line: AI won’t replace physicians—but if you’re a radiologist, you should be concerned.
Let’s examine that claim through different lenses.
Assume it’s true. Suppose other physicians are safe from AI, and radiologists are not. In that case, perhaps young physicians should hesitate before entering radiology and consider switching to a “safer” specialty. We all know the standard arguments: radiology is image-based; images are digital; AI excels at digital pattern recognition; and it will only continue to improve until it rivals—or surpasses—human image interpretation.
But we also know the counterarguments. Radiologists do far more than read images. We perform image-guided procedures, synthesize findings across modalities, integrate lab data and clinical context, consult with clinicians and patients, and take responsibility when things go wrong. We are someone to call when there’s uncertainty—and someone to sue when something is missed. And despite the hype, AI is not replacing us today. Objectively, demand for radiologists is near an all-time high, with shortages nearly everywhere.
Now let’s consider the opposite assumption: that the original claim is false—that radiology is not uniquely vulnerable. In that scenario, a trainee might switch out of radiology only to land in another specialty that ultimately faces the same disruption.
If AI replaces physicians, in whole or in part, what reason do we have to believe this will be exclusive to radiology?
One of my interventional radiology colleagues recently remarked that AI-driven vascular robots will almost certainly perform many of the procedures IRs do today. A general surgeon friend described autonomous robotic cholecystectomies already performed successfully on animals. I personally wouldn’t be opposed to enrolling in an AI-based primary care platform that handles many routine tasks my current PCP performs.
It’s not hard to imagine AI systems taking a complete medical history, running ventilators, placing central lines, interpreting EKGs, ordering and analyzing labs, performing variations of a physical exam, completing colonoscopies and bronchoscopies, evaluating pathology specimens, conducting neurological exams, operating with extreme precision on the brain, heart, and lungs, performing C-sections, fixing fractures, replacing joints, performing cataract surgery, administering chemo- or immunotherapy regimens, managing autoimmune diseases, diagnosing and treating infectious diseases, robotic systems successfully running codes, and much more.
My point is simple: no area of medicine can confidently claim immunity from AI. Perhaps AI won’t replace physicians outright—but even widespread augmentation could mean fewer physicians are needed. In a system currently facing shortages, that could eventually flip to surplus.
So, if parents and colleagues are truly worried about obsolescence, perhaps they shouldn’t be steering trainees toward other medical specialties—but out of medicine entirely. But then where? Law? Programming? Finance? An MBA? Are these professions immune?
You see the problem, and it extends far beyond radiology.
The question of radiology obsolescence is really a question about the future of white-collar work itself—any profession rooted in human knowledge, analysis, or judgment. If that’s the concern, maybe the safest advice isn’t “don’t be a radiologist,” but “enter a trade or own an experiential business such as a resort.” Though even those may not be safe forever. Perhaps the best hedge is owning power plants.
The takeaway is this: if you’re young and worried about radiology, you should probably be worried about nearly every other career resting on human judgement too. Radiology may be in the spotlight, but if AI replaces us, we certainly won’t be the only ones.
So, what can be done?
One colleague joked about “obsolescence insurance.” While such a policy isn’t financially realistic, the concept is useful: what can you do to hedge against irrelevance?
For radiologists—especially trainees and those early in their careers—one answer is skill diversification. I never felt comfortable exclusively practicing breast radiology as a fellowship-trained breast radiologist. While there are lifestyle advantages to such a narrow practice—no call, no weekends, no holidays—it also concentrates risk, particularly when your livelihood depends heavily on one modality that AI targets aggressively: the mammogram.
If screening mammography becomes even partially autonomous, the implications for the breast imaging workforce could be profound. Yes, we also interpret diagnostic mammograms, ultrasound and MRI and perform biopsies and procedures that are harder to automate—but screening keeps the lights on.
That’s why I believe early breast imagers should consider participating in general call or dual-subspecializing, as I did in nuclear radiology. Since graduating fellowship, and for many years reading exclusively breast and nuclear radiology, I broadened further by taking general call. You can always narrow your scope later—but staying broad into your early- and mid-radiology career may function as a form of obsolescence insurance.
This principle applies across subspecialties. Going 100% into any single subspecialty area increases vulnerability if demand for that niche falls. On the other hand, it’s possible that the most specialized experts become the last ones standing—overseeing AI systems across multiple hospitals. I would place my own bets that the opposite happens, and broadly trained radiologists augmented by AI become the most valuable. No one knows.
When uncertainty reigns, history is instructive. We’ve faced workforce disruptions before. AI is different—yes—but that doesn’t mean today’s predictions will be accurate.
When I entered medical school nearly two decades ago, I was also told not to go into radiology—but for different reasons. The job market was terrible. Radiologists couldn’t find positions, and those that existed were often undesirable. I took a leap of faith, knowing I might struggle to find work. I was never recruited as a trainee; I had to actively market myself.
And yet, just as I finished fellowship, the market flipped almost overnight. Demand exploded—and it hasn’t slowed since. The naysayers were wrong.
I was also told by oncologists not to pursue breast imaging because cancer treatments were becoming so effective that screening would soon be unnecessary. I’m still waiting for a cure for metastatic breast cancer. Outcomes have improved—but early detection remains critical.
The lesson isn’t that radiology—or any specialty—is immune to obsolescence. It isn’t. But living in constant fear of predictions that may or may not materialize is not the most adaptive response.
What will matter most in the future are timeless skills: determination, adaptability, intelligence, perseverance, hard work, responsibility, humility, technical ability, and a relentless willingness to learn. Radiology cultivates all of these.
Predictions about the future will always sound confident—right up until they’re wrong. Radiology is not immune to change, but neither is any profession built on human judgment. If radiology truly becomes obsolete, it won’t be because of AI alone—and we won’t be the only ones affected. Until then, fear is a poor career strategy. Adaptability, breadth, and a commitment to excellence are better ones. Those qualities have carried radiology through every prior disruption, and they may just carry us through this one as well.
